Using Windows Performance Toolkit to find System Issues in Vista/Win2k8/Win7

Windows 7 RC1 just came out. I am a TechNet subscriber, so I wanted to try it out. I have an old (2005) Dell desktop, 2.8 GHz, 2 GB ram, 160 GB drive box. 3.7 rating for Vista (because of the Graphics card mostly, would be 4.4 otherwise – not too bad, even for being kind of an old box). It has been sitting in the basement since I moved into my new place in October, doing nothing really. I use Mac full time at home, so it just sits.

A few times I have tried to get Windows Vista running smooth on it, Media Center, or just a file server,etc. Thing is, it was just flaking out. I knew it was a hardware issue, but figured it might be the CPU fan, or overheating, etc. Vista installed fine, but as I was using it, I would see just hang-ups, lockups. Not BSOD’s, but it would just hang, for 30 seconds, 1 minute, and then come back. WTF?

Nothing in the Reliability monitor, nothing I could see in event logs, etc. I rebooted, did Windows Memory test, nothing there. If you go into Computer Management, you will see Performance, then Data Collector Sets and Reports, Monitoring Tools. You can set it up to run a test on metrics of your system and it will give you a report

image

I did this, and everything was ok. BUT… Avg Disk Length Queue was > 2 – red flag. Disk issues. But I wanted to know more. So I started digging around, and there is a Windows Performance Toolkit you can download. Here is another good site going into detail about the WPT.

So I fire up cmd line (as admin! – start->run, cmd ctrl+shift+enter), and run

xperf -providers K

to see what providers are available for the Kernel flags. IOTrace looks like something I want, so I then run

xperf -on IOTrace

and let it run. I go and open/close things, play around, see if I can replicate the issue. Once I feel I have, I want to stop and analyze the trace. You need to stop it and output to a file using this command:

xperf -d iotrace.etl

Side note: Files are named ETL. Coming from a BI background, this makes my world explode, since it has nothing to do with Extract, Transform, and Load

Now that my trace is done, time to analyze:

xperfview iotrace.etl

And you get some awesome stats like this:
image

Although I didn’t save my stats from my tests that showed the bad IO, what I saw were just gaps in the graphs, glitches in The Matrix. Time missing. Something is really bad here. So I did the drive error checking in Vista:

image

And when that ran, after reboot, it got to 11% and croaked. Bad drive. So I went and bought a new 500 GB SATA drive and loaded it up, and I am running Windows 7 now. Pretty sweet.

After all this fun spelunking into Windows performance, it also got me thinking about things, like running these detailed traces on SQL Server boxes or other servers on intervals, and saving them somehow, reporting on the data. The IOTrace is just one of hundreds of traces, that you can then auto analyze. I know that there are perfmon tools but there are some added benefits to xperf that you can you utilize, and I am glad I learned more about it and put it to use, just another tool for the sysadmin tool belt.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s